CRN Stage-II Dataset: Transport-derived decision bias on 1064 HCP connectomes

Models of perceptual decision-making describe how noisy evidence is accumulated to a commitment threshold, yet what determines the quality of evidence entering the accumulator remains largely unspecified. We address this gap with a two-stage computational architecture in which a first stage of coherent network transport, formalised via a Lindblad master-equation framework employed as a functional proxy for wave-like network exploration rather than a literal quantum-mechanical claim, produces a scalar evidence signal for each of 1,064 structural connectomes from the Human Connectome Project. This signal is then fed into two established Stage-II accumulators: a drift-diffusion model, where it sets the starting point of the decision variable, and a Wong–Wang attractor network, where it biases the input current near a symmetry-breaking bifurcation. Both models detect a significant paired advantage of the coherent-optimal over the classical transport condition (attractor-network Wilcoxon z = −28.07, p < 10⁻¹⁷³; drift-diffusion z = −8.79, p < 10⁻¹⁸). A between-subject dose–response confirms that individuals with a larger transport advantage show stronger reaction-time modulation (Pearson r = −0.645; partial r controlling for graph density = −0.606). Mediation analysis on the full cohort shows that graph topology affects commitment dynamics exclusively through this scalar evidence signal, with no residual direct path. The two coupling mechanisms yield opposite predictions about error speed: the drift-diffusion model predicts that correct responses are faster than errors, whereas the attractor network predicts the reverse. Together with three further experimentally discriminable signatures in the reaction-time distribution, these predictions provide a falsifiable bridge between structural transport properties and behavioural decision dynamics that can be tested with forced-choice reaction-time data on the same cohort. Related deposits: HCP reproducibility bundle (doi:10.5281/zenodo.18519173), Drosophila mushroom body dataset (doi:10.5281/zenodo.18697116), Mouse cortex applicability boundary (doi:10.5281/zenodo.18968563).